Two-Way Tables

When statisticians want to study the relationship between two categorical variables simultaneously, they organize the data into a two-way table, also called a contingency table or cross-tabulation. For example, a researcher might ask whether a student's preferred study method (solo, group, or online) is related to their grade category (A, B, or C). Each cell in the table records the count of individuals who fall into both categories at once, making it easy to spot patterns that would be invisible if the variables were examined separately. Two-way tables are a cornerstone of exploratory data analysis. They allow us to compute marginal frequencies (totals along the edges of the table), joint frequencies (counts inside the table), and conditional frequencies (proportions within a row or column). These three perspectives let us answer very different questions from the same dataset, so mastering how to read and build these tables is an essential skill in Statistics. Understanding two-way tables also lays the foundation for more advanced statistical tests, most notably the chi-square test of independence, which formally determines whether the two categorical variables are associated or whether any apparent pattern could be explained by random chance. Even before reaching that test, two-way tables give you an immediate, intuitive window into the structure of bivariate categorical data.

Key takeaways
  • Identify the Two Categorical Variables
  • Create the Table Shell
  • Fill in Joint Frequencies
  • Compute Marginal Frequencies
  • Convert to Relative Frequencies
  • Detect Association

Identify the Two Categorical Variables

Label one variable as the row variable and the other as the column variable. There is no statistical rule for which goes where; choose the arrangement that makes the table easiest to read. For example: rows = Gender (Male, Female), columns = Preferred Sport (Soccer, Basketball, Tennis).

Create the Table Shell

Draw a grid with one row for each category of the row variable plus a 'Total' row, and one column for each category of the column variable plus a 'Total' column. This gives an (r+1) × (c+1) grid, where r is the number of row categories and c is the number of column categories.

Fill in Joint Frequencies

Count how many individuals belong to each combination of categories and enter those counts in the interior cells. These counts are called joint frequencies. For instance, if 34 students are Male AND prefer Soccer, write 34 in the Male row, Soccer column.

Compute Marginal Frequencies

Add across each row to get the row totals, and add down each column to get the column totals. These edge totals are called marginal frequencies because they appear in the margins of the table. The bottom-right corner cell is the grand total n, which equals the sum of all row totals and also the sum of all column totals.

Convert to Relative Frequencies

Divide every cell (joint or marginal) by the grand total n to obtain joint relative frequencies and marginal relative frequencies. These proportions always sum to 1. Dividing each cell by its row total gives row conditional relative frequencies; dividing by the column total gives column conditional relative frequencies. Comparing conditional distributions across rows or columns reveals whether an association exists between the two variables.

Detect Association

If the conditional distribution of the column variable is roughly the same for every row, the two variables appear independent. If the conditional distributions differ noticeably across rows, an association is present. For example, if 70% of Males prefer Soccer but only 30% of Females do, gender and sport preference are associated.

The Key Equation

Conditional Relative Frequency=fijfi⋅\text{Conditional Relative Frequency} = \frac{f_{ij}}{f_{i\cdot}}
  • f_{ij} — Joint frequency (count): The number of individuals in row category i AND column category j.
  • f_{i\cdot} — Row marginal frequency (count): The total number of individuals in row category i, obtained by summing across all column categories.
  • \frac{f_{ij}}{f_{i\cdot}} — Row conditional relative frequency (dimensionless (proportion)): The proportion of individuals in row i who also belong to column category j. Values in a row sum to 1.

Worked Example

Problem. A school surveyed 200 students about their lunch choice (Hot Meal or Packed Lunch) and their year group (Year 10 or Year 11). The results: 70 Year-10 students chose Hot Meal, 30 chose Packed Lunch; 50 Year-11 students chose Hot Meal, 50 chose Packed Lunch. (a) Construct the two-way table with marginal totals. (b) Find the joint relative frequency for Year-11 students who chose Packed Lunch. (c) Find the conditional relative frequency of choosing Hot Meal given the student is in Year 10. (d) Does an association appear to exist between year group and lunch choice?

50200=0.25and70100=0.70\frac{50}{200} = 0.25 \quad \text{and} \quad \frac{70}{100} = 0.70

Answer. Joint relative frequency (Year 11, Packed Lunch) = 0.25; Conditional relative frequency (Hot Meal | Year 10) = 0.70; an association exists.

Figure
Diagram
Two-Way Table: Lunch Choice by Year GroupYear GroupHot MealPacked LunchTotalYear 107030100Year 115050100Total12080200Row conditional frequencies (Hot Meal):Year 10: 70/100 = 70% Year 11: 50/100 = 50%Difference in conditional distributions suggests an association.
Origins · 1 minKarl Pearson and how this was discovered
KP
Karl Pearson
1857–1936 · British

Karl Pearson formalized the chi-square test (1900), which is the primary hypothesis test used with contingency tables. He also coined the term 'standard deviation' and founded the first university statistics department in the world at University College London. His work gave statisticians the tools to decide whether patterns seen in two-way tables are statistically significant or just the result of random variation. Pearson was so passionate about statistics that he ran biometric laboratories funded partly by eugenics societies, a deeply controversial chapter in his legacy that modern statisticians reflect on critically. Despite the ethical concerns surrounding some of his applications, his mathematical contributions remain foundational to the field.

  • Frequency Tables for Categorical Data — Two-way tables extend single-variable frequency tables to two variables; understanding one-way frequency tables is the direct prerequisite.
  • Relative Frequency and Proportions — Converting counts to proportions is central to interpreting two-way tables through joint and conditional relative frequencies.
  • Chi-Square Test of Independence — The chi-square test uses two-way tables to formally test whether two categorical variables are statistically independent.
  • Conditional Probability — Conditional relative frequencies in two-way tables are the empirical (data-based) version of conditional probability P(A|B).
  • Bar Charts for Categorical Data — Segmented and side-by-side bar charts visually display the same conditional distributions found in two-way tables, bridging numerical and graphical analysis.